Post on 26-May-2020
Routing & Protocols
1
Paul Trainacisco Engineering
2
Paul Traina / INET '95Developing Countries Workshop
Today's Talk
�Terminology�Routing�Static Routes�Interior Gateway Protocols�Exterior Gateway Protocols�Building an ISP network
3
Paul Traina / INET '95Developing Countries Workshop
Terminology
�network number�prefix�mask (or length)
4
Paul Traina / INET '95Developing Countries Workshop
Static routeshand configured routing
�tell the router which way to send packets�based upon final packet destination
5
Paul Traina / INET '95Developing Countries Workshop
Static routes
�ip route 10.0.0.0 255.0.0.0 serial 3�ip route 131.108.0.0 255.255.0.0 171.65.3.4
6
S3
171.65.3.4
Paul Traina / INET '95Developing Countries Workshop
TerminologyInterior Gateway Protocol (IGP)
�RIP, IGRP, HELLO, OSPF�Primary goal is optimal connectivity�Strong distance metrics�May not have good administrative controls
7
Paul Traina / INET '95Developing Countries Workshop
TerminologyDistance vector protocols
�listen to neighboring routers�install routes in table, lowest distance wins�advertise all routes in table�very simple�very stupid
8
Paul Traina / INET '95Developing Countries Workshop
TerminologyDistance vector protocols
9
A
B
C
D
E
F
G
H IA 1B 1C 1G supH 1
A 2B 2C 2D 1E supF 1G 1H 1I1
Paul Traina / INET '95Developing Countries Workshop
TerminologyLink state protocols
�information about adjacencies sent to all routers�each router builds a topology database�a "shortest path" algorithm is used to find best route�converge as quickly as databases can be updated
10
Paul Traina / INET '95Developing Countries Workshop
TerminologyLink state protocols
11
A
B
C
D
E
F
G1 2
A - 1 - G - 2 - D
H I3
router 1A, B, C, G, H
router 3H, I
router 2D, E, F, G, I
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsRouting Information Protocol (RIP)
�IP only�distance vector protocol�slow convergence�does not carry mask information�reasonably simple design & configuration�does not scale (maximum 15 hops)�poor metrics (hop-count)
12
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsInterior Gateway Routing Protocol (IGRP)�IP only�distance vector protocol�slow convergence (like RIP)�does not carry mask information (like RIP)�very simple design & configuration
�powerful proprietary metric�load sharing across diverse links
13
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsThe IGRP metric
»always get optimal routing�metric vector, not single value
�bandwidth�delay�hops�reliability�loading
14
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsEnhanced IGRP
�multi-protocol (IP, IPX, Appletalk)�fast convergence (like OSPF)�very simple design & configuration (like IGRP)
�IGRP metric�allows load sharing across diverse links
15
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsEnhanced IGRP
�distance vector based protocol�NOT a Bellman-Ford protocolUses "dual" algorithm�alternative to OSPF & I-ISIS�can be bandwidth intensive on slow links
16
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsIntegrated IS-IS (I-ISIS)
�multi-protocol (CLNP, IP, IPX, ...)�link state protocol�fast convergence�design and architecture moderately complex�configuration may be simple
17
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsOpen Shortest Path First (OSPF)
�IS - IS = 0
18
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsOpen Shortest Path First (OSPF)
�IP only�link state protocol�fast convergence�design and architecture very complex�configuration can be simple
19
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsWhich to use?
�Your interior network is actually VERY simple.�Your IGP should only carry your routes and your direct customers'
20
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtcolsProblems with "classic" protocols
�slow convergence�count to infinity�no mask information
�no CIDR�no VLSM�no subnet 0
21
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsSlow convergence
�advertisement period�entire routing table dumped every n seconds
�timeout period�usually 3 times advertisement period
�RIP values are normally 30 and 90 seconds!
22
Paul Traina / INET '95Developing Countries Workshop
1 (3 hops)
Interior Gateway ProtocolsCount to infinty problem
23
1
A B C
2 34
1 (1 hop) 1 (2 hops)
1
A B C
2 34
1 (3 hops)1
A B C
2 34
1 (3 hops)1 (4 hops)
route cleared
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsCount to infinity: split-horizon
�Don't feed selected route back to source
�no feedback on source interface�no feedback to source neighbor
24
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsCount to infinity: split-horizon
25
1 (3 hops)
1
A B C
2 34
1
A B C
2 34
1 (3 hops)1
A B C
2 34
route cleared
1 (1 hop) 1 (2 hops)
route cleared
route cleared
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsCount to infinity: hold-down
�Split horizon not sufficient!�Holddown period
�interval during which "less attractive" updates are ignored
26
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsCount to infinity: hold-down
27
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsThe universal rule
�You will always trade bandwidth for speed of convergence
28
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsOSPF configuration
�myth�OSPF is hard to use
�reality:�router ospf 1network 192.111.107.0 0.0.0.255 area 0
29
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsOSPF operation
�every OSPF router sends out 'hello' packets�hello packets used to determine if neighbor is up�hello packets are small easy to process packets�hello packets are sent periodically (usually short interval)
30
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsOSPF operation
�once an adjacency is established, trade information with your neighbor�topology information is packaged in a "link state announcement"�announcements are sent ONCE, and only updated if there's a change
�(or every 45mins...)
31
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsOSPF operation
�change occurs�broadcast change�run SPF algorithm�install output into forwarding table
32
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway Protocolsmaking OSPF scale
�each link transition causes a broadcast and SPF run�OSPF can group routers to appear as one single router�OSPF areas
33
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsOSPF areas (before)
34
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsOSPF areas (after)
35
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsOSPF areas - partitioning
36
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsOSPF areas - partition repair
37
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsOSPF areas
�rule of thumb:no more than 150 routers/area�reality:no more than 500 routers/area�backbone "area" is an area�proper use of areas reduce bandwidth & CPU utilization
38
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsEIGRP operation
�design goals were�make it as fast as OSPF & IS-IS�make it trivial to configure�easy migration from IGRP
39
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsEIGRP operation
�router eigrp 1network 192.108.0.0 mask 255.255.0.0
40
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway ProtocolsEIGRP operation - caveats
�nothing is for free�EIGRP works best on high speed links�EIGRP doesn't scale well in high-meshed frame-relay networks
�star networks OK
41
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway Protocolssummarization
42
�classful routing protocols naturally summarize to network numbers at boundaries
131.108
150.196150.196.4.0
131.108.3.0
.1.0
.2.0
.3.0
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway Protocolssummarization
�classless routing protocols summarize at arbitrary bit boundaries
43
131.108.4.32/29131.108.3.64/27
131.108.0.0/16
131.108.4.0/24
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway Protocolsroute filtering
�pseudo-security (bad idea!)�low bandwidth links�eliminate unnecessary information
44
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway Protocolsroute filtering
45
B
131.108/16150.136/1644/80/0A
C
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway Protocolsredistribution
46
�you run OSPF�your neighbor runs RIP
RIP
OSPF
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway Protocolsredistribution
�run RIP on their interface�router ripnetwork 192.111.107.0�configure OSPF to redistribute RIP�router ospf 1network 135.111.104.0 0.0.0.255 area 0redistribute rip metric 10
47
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway Protocolsredistribution
�bi-directional redistribution MUST be filtered!
48
Paul Traina / INET '95Developing Countries Workshop
Interior Gateway Protocolsredistribution�router ripnetwork 192.111.107.0�router ospf 1network 135.111.104.0 0.0.0.255 area 0redistribute rip metric 10distribute-list 1 out rip�access-list 1 permit 192.111.107.0 0.0.0.255
49
xxJob Number Goes Here
Exterior routing
50
Paul Traina / INET '95Developing Countries Workshop
Exterior routing
�Terminology�What is exterior routing?�Routing protocols�Overview of BGP�Putting it all together�Further information
51
Paul Traina / INET '95Developing Countries Workshop
TerminologyAutonomous System�A set of networks sharing the same routing policy.�Internal connectivity�One contiguious unit�Identified by "AS number"�Examples
�service provider�multi-homed customer�anyone needing policy discrimination
52
Paul Traina / INET '95Developing Countries Workshop
TerminologyExterior routes
�Routes learned from other autonomous systems
53
Paul Traina / INET '95Developing Countries Workshop
TerminologyExterior Gateway Protocol
�egp vs EGP�EGP, BGP, IDRP�Primary goal is to provide reachabilityinformation outside administrative domain�Secondary goal is administrative control�Metrics may be arbitrary or weak
54
Paul Traina / INET '95Developing Countries Workshop
TerminologyNatural network mask
�Classful mask�Class A = 8 bits
�networks 1...127�Class B = 16 bits
�networks 128.0...191.255�Class C = 24 bits
�networks 192.0.0...223.255.255
55
Paul Traina / INET '95Developing Countries Workshop
TerminologyDMZ network
�de-militarised zone�area between North and South Korea�shared network between ASs
�before, neither AS carried it in IGP�now, both carry it in IGP
56
Paul Traina / INET '95Developing Countries Workshop
TerminologyDMZ network
57
DMZ networks
Paul Traina / INET '95Developing Countries Workshop
Why do we need exterior routing?Why not make entire internet a single cloud?
�separate policy control�filtering on networks doesn't scale well�service provider selection given multiple choices�everything must scale to hundreds of thousands of routes
58
Paul Traina / INET '95Developing Countries Workshop
Exterior Routing
�static routes�multiple IGP instances�OSPF inter-domain routing�EGP�IDRP�BGP version 4
59
Paul Traina / INET '95Developing Countries Workshop
Exterior RoutingStatic routes
�no path information�very versatile�low protocol overhead�high maintenance overhead�very very very bad convergence time
�requires manual configuration
60
Paul Traina / INET '95Developing Countries Workshop
Exterior RoutingMultiple IGPs with route leaking�Run an instance of an IGP at each site for local routing�Run a backbone IGP at each border router�redistribute local IGP into backbone IGP�redistribute backbone IGP into local IGP (or default)�backbone routers share common administration
61
Paul Traina / INET '95Developing Countries Workshop
Exterior RoutingMultiple IGPs with route leaking
62
customer
service providerIGRP 109
OSPF 690
RIP run over wire
rip default redistributed intocustomer's IGP
RIP routes learned fromcustomer redistributedinto service provider's IGPafter filtering
Paul Traina / INET '95Developing Countries Workshop
Exterior RoutingMultiple IGPs with route leaking
�backbone IGP�router ospf 690network 129.119.0.0 0.0.255.255 area 0redistribute rip metric 5distribute-list 1 rip out
�local IGP�router igrp 109network 131.108.0.0ip default-network 140.222.0.0
63
Paul Traina / INET '95Developing Countries Workshop
Exterior RoutingOSPF inter-domain routing
�Route leaking formalised for one protocol�OSPF tag carries originating AS
�limited policy control�only have 32 bit OSPF tag�OSPF tag contains originating AS
64
Paul Traina / INET '95Developing Countries Workshop
Exterior RoutingExterior Gateway Protocol
�historical protocol�obsolete�assumes a central core�no transit service except via core
65
Paul Traina / INET '95Developing Countries Workshop
Exterior RoutingExterior Gateway Protocol (historical)
�RIP by any other name�fancy "hello dance"�periodic update protocol�entire routing table sent with each update�no metric
�everything is one hop from core
66
Paul Traina / INET '95Developing Countries Workshop
Exterior RoutingExterior Gateway Protocol
67
core
109 110
�AS 110 may not advertise AS 109 to core
Paul Traina / INET '95Developing Countries Workshop
IDRP (future expansion path)Inter-domain routing protocol
�IDRP is an almost identical clone of BGP-4�IDRP is multi-protocol
�IP�CLNP�IPX
�For purposes of this talk: g/BGP-4/s//IDRP/g
68
Paul Traina / INET '95Developing Countries Workshop
BGP-4Border Gateway Protocol version 4
�carries external routes only�uses reliable transport mechanism (TCP)�not a periodic routing protocol�allows limited policy selection�AS path insures loop free routing�"best path" determined at AS granularity
69
Paul Traina / INET '95Developing Countries Workshop
BGP peer relationshipsExternal BGP
�neighbor is in a different AS�neighbors share a common network
70
AS 110
AS 109
Paul Traina / INET '95Developing Countries Workshop
BGP peer relationshipsInternal BGP
71
�neighbor in same AS�may be several hops away�full neighbor mesh required
Paul Traina / INET '95Developing Countries Workshop
Common BGP networksStub customer
72
�BGP only at border�default to border
Paul Traina / INET '95Developing Countries Workshop
Common BGP networksMulti-homed customer
�Internal BGP used with IGP�IBGP only between border gateways�Only border gateways speak BGP�Synchronization with IGP required�May use one IGP for exterior routes, and another for internal nodes
�exterior routes must be redistributed into IGP
73
Paul Traina / INET '95Developing Countries Workshop
Common BGP networksMulti-homed customer
74
Paul Traina / INET '95Developing Countries Workshop
Common BGP networksService provider
�Internal BGP used to carry exterior routes
�IGP carries local information only�Full mesh required if no IGP synchronization
75
Paul Traina / INET '95Developing Countries Workshop
Common BGP networksService provider
76
Paul Traina / INET '95Developing Countries Workshop
Common BGP networksService provider confederation
77
�A group of service providers�Multiple connectivity points
�multi-exit discriminator useful�Not a special case
Paul Traina / INET '95Developing Countries Workshop
The BGP protocolUpdate messages
�withdrawn routes�attributes�advertised routes
78
Paul Traina / INET '95Developing Countries Workshop
Update messagesNetwork reachability information�prefix length
�number of significant bits�network prefix
�0 to 4 bytes�Example:
�131.108/16�131.108.0.0 255.255.0.0
�193/8�193.0.0.0 255.0.0.0
79
Paul Traina / INET '95Developing Countries Workshop
Update messagesAttributes
�AS path�next hop�origin�local preference�multi-exit discriminator�atomic aggregate�aggregator
80
Paul Traina / INET '95Developing Countries Workshop
AS pathAS sequence
�a list of AS's that a route has traversed�109 200 690 1755 1883
81
Paul Traina / INET '95Developing Countries Workshop
AS pathAS sequence
82
193.0.33/24
193.0.35/24
193.0.34/24
1880
1881
1882193.0.33/24 1880 1881193.0.34/24 1880193.0.35/24 1880 1882
Paul Traina / INET '95Developing Countries Workshop
AS pathAS set
�path traversed one or more members of a set
�{1880,1881,1882}
83
Paul Traina / INET '95Developing Countries Workshop
AS pathAS set
84
193.0.31/24
193.0.32/24
193.0.33/24
1880
1881
1882
193.0.32/22{1880,1881,1882}
Paul Traina / INET '95Developing Countries Workshop
AS pathSets and sequences combined
�local aggregation�109 200 690 1755 {1881,1882,1883}
�regional aggregation�109 200 690 {1755,1881,1882,1883,...}
85
Paul Traina / INET '95Developing Countries Workshop
BGP path selectionBGP maintains multiple "feasable" paths to a destination
�fast convergence�routing based upon preferences�Example:
�131.108/16 may be reached via AS path 690 200 109 or via AS path 690 1340 109
86
Paul Traina / INET '95Developing Countries Workshop
BGP path selection algorithmInitial route determination
�do not consider path if no next hop route�largest weight
�local to router�highest local preference
�global within AS�shortest AS path
87
Paul Traina / INET '95Developing Countries Workshop
BGP path selectionTie breaking
�multi-exit discriminator�only considered if AS paths identical
�external routes�best IGP metric to next hop�highest IP address
88
Paul Traina / INET '95Developing Countries Workshop
Policy Control
�distribute list�filter individual networks
�filter list�filter by AS path
�route maps�general policy control and tuning
89
Paul Traina / INET '95Developing Countries Workshop
More informationTechnical information on BGP
�RFC-1772�application of the Border Gateway Protocol
�RFC-1771�BGP-4 protocol reference document
�RFC-1745�BGP <-> OSPF interaction
90
xxJob Number Goes Here
Building an Internet
91
Paul Traina / INET '95Developing Countries Workshop
Putting it all togetherGeneral philosophy
�Your network is going to grow at an exponential rate!�Design to scale...but be prepared to reorganize from scratch�Don't be afraid of change!
�Most network redesigns are only configuration changes
92
Paul Traina / INET '95Developing Countries Workshop
Putting it all together
�Requirements for IGPs for backbones�IGP connects your backbone together, not your client's routes�Must
�converge quickly�Should
�carry netmask information
81
Paul Traina / INET '95Developing Countries Workshop
Putting it all togetherconnecting to a customer�static routes
�you control directly�no route flaps�no packets to be charged
�shared routing protocol or leaking...�you MUST filter your customers info�route flaps
�BGP for multi homed customers
82
Paul Traina / INET '95Developing Countries Workshop
Putting it all togetherbuilding your backbone
�keep it simple�redundancy is good, but expensive�use an IGP that carrys mask information�use an IGP that converges quickly�use OSPF, ISIS, or EIGRP
83
Paul Traina / INET '95Developing Countries Workshop
Putting it all togetherconnecting to other ISPs
�Use BGP-4�advertise only what you serve�take back as little as you can
84
Paul Traina / INET '95Developing Countries Workshop
Putting it all togetherthe internet exchange
�long distance connectivity is expensive�connect to several providers at a single point
85
Paul Traina / INET '95Developing Countries Workshop
Internet exchanges - FIXFederal internet exchange (historical)
�dumb ethernet connecting a group of service providers
86
Paul Traina / INET '95Developing Countries Workshop
Internet exchanges - FIXFederal internet exchange
87
�single primary media all systems share�secondary media may be shared by a subset of systems to reduce load on primary media
Paul Traina / INET '95Developing Countries Workshop
Non-Internet exchange - CIXCommercial internet exchange (historical)
�actually a one-router transit AS�CIX clients only receive best path as determined by CIX router
88
CIX
Paul Traina / INET '95Developing Countries Workshop
Internet exchanges - d-GIXDistributed global internet exchange
89
�emulates a single ethernet
Paul Traina / INET '95Developing Countries Workshop
Internet exchanges - d-GIXDistributed global internet exchange
�share the cost of high speed lines�single virtual level-2 media
�bridges, not routers, connect the link access points
�bridge table entries are static�don't need spanning tree�mac address filtering used
90
Paul Traina / INET '95Developing Countries Workshop
Internet exchanges - d-GIXDistributed global internet exchange
�the GIX itself still has no routing policy�in that case, how do you pay for it?�the GIX does have connectivity policy
�charge for MAC address filters (source/destination filtering)
91
Paul Traina / INET '95Developing Countries Workshop
Internet exchanges - multi-NAPMultiple-media network access point
92
NetworkAccessPoint
Frame Relay
ATM
local ethernet
Paul Traina / INET '95Developing Countries Workshop
Internet exchanges - multi-NAPMultiple-media network access point
�Problem:�How do you allow one NAP client to connect via Frame Relay and another customer connect via ATM?
�Answer:�Don't do this! Extend the NAP and keep it policy free.
93
Paul Traina / INET '95Developing Countries Workshop
Interenet exchanges - multi-NAPMultiple-media network access point
�NAPs and IXs need to be policy free�Routers implicity have an 'advertise only what you use' policy.�If routers are used, NAP becomes a transit AS, not an "IX," and clients of the NAP are limited by the NAP's route selection policy.
94
Paul Traina / INET '95Developing Countries Workshop
More information
�Original GIX proposal�ftp://ftp.ripe.net/ripe/docs/ripe-082.ps�ftp://ftp.ripe.net/ripe/drafts/gix15jun.txt
�d-GIX - distributed global internet exchange
�ftp://ftp.ripe.net/ripe/drafts/d-gix-proposal.ps
95
Paul Traina / INET '95Developing Countries Workshop
Routing registriesWhat are they?
�database containing�route prefix/origin autonomous system�autonomous system/connectivity policy
�RIPE-181 aka RC-1786
96
xxJob Number Goes Here
Classless routing
97
Paul Traina / INET '95Developing Countries Workshop
Why CIDR?
�IP route advertisements have been growing exponentially.�Class A networks are too big�Class C networks are too small�Only 65534 class B networks available
98
Paul Traina / INET '95Developing Countries Workshop
Routing Table Growth
99
87 88 89 90 91 92 93 94 95 960
10000
20000
30000
40000
50000
60000
Paul Traina / INET '95Developing Countries Workshop
Why CIDR?Classful networks mis-sized
�Class A networks are too big�not desirable because of connectivity constraints
�Class B address space is depleted�Class C networks are useful only for small customers
�large gap between "C" customer and "B" customer
100
Paul Traina / INET '95Developing Countries Workshop
Classless routingCIDR at the service provider level
�Service provider given CIDR blocks by numbering authority�Example:
�198.24/15 == 512 class "C" nets�Service provider advertises only a summary route for CIDR block to neighboring providers, not 512 separate class "C" routes.
101
Paul Traina / INET '95Developing Countries Workshop
Classless routingThe client interface
�Partition local CIDR block and assign to customers�Example:
�198.24.62/23 == 2 "C" nets�198.24.192/18 == 64 "C" nets�198.24.61/24 == 1 "C" net
102
Paul Traina / INET '95Developing Countries Workshop
Classless routingDo's and don'ts
�Don't assign blocks smaller than class "C" sized networks without prior agreement from customers
�most hosts & routing protocols are not classless
�Do help customers use their address space wisely!
103
Paul Traina / INET '95Developing Countries Workshop
Classless routingDo's and don'ts
�Do give customers enough address space for what they need�Do parition your CIDR block to provide for customer growth
�get the tree program�understand RFCs 1519 and 1219
104
Paul Traina / INET '95Developing Countries Workshop
Classless routingDo's and don'ts
�Don't be afraid of "holes" when aggregating�Longest match routing means "he who has the longest prefix wins"
105
Paul Traina / INET '95Developing Countries Workshop
Classless routingGetting the most out of your allocation
�It's natural, but inefficent to subnet on 8 bit boundaries
�131.108.1 = subnet 1�131.108.2 = subnet 2�131.108.3 = subnet 3
�254 subnets with up to 254 hosts persubnet out of a 16 bit address allocation
106
Paul Traina / INET '95Developing Countries Workshop
Classless routingThere are NO NETWORK NUMBERS!!!
�...just address space prefixes�131/8�131.0/12�131.108/16�131.108.5/24�131.108.5.32/29�131.108.5.33/32
107
Paul Traina / INET '95Developing Countries Workshop
Classless routingThere are NO SUBNET MASKS!!!
�It's no longer a mask, just a prefix length�There can be no '0' holes in the mask�/16 = 255.255.0.0�/32 = 255.255.255.255�/14 = 255.252.0.0�/0 = default = 0.0.0.0
108
Paul Traina / INET '95Developing Countries Workshop
Classless routingGetting the most out of your allocation
�Unnumbered serial links�Variable length subnet masks�Small ethernet
�28 bit mask = 14 hosts�Larger ethernet
�26 bit mask = 62 hosts�VLSM allocation rules are the same as CIDR allocation
109
Paul Traina / INET '95Developing Countries Workshop
Classless routingrestrictions removed
�no such thing as a "subnet" anymore�subnet 0 is no longer special�all 1's subnet is no longer special�no such thing as a disconnectedsubnet
110
Paul Traina / INET '95Developing Countries Workshop
Classless routingMickey Mouse topology is OK
111
131.108.0/17 131.108.128/17
192.111.107/24
Paul Traina / INET '95Developing Countries Workshop
Classless routingPlan for entropy
�What is your policy when customers move to a different service provider?
�do you own the numbers in the CIDR block?�will new service provider supply more specific routing information?
112
Paul Traina / INET '95Developing Countries Workshop
Classless routingAllocate addresses efficiently!
�you don't get very many�what happens as organizations grow?�what happens when your customers lie to you?
113
Paul Traina / INET '95Developing Countries Workshop
More informationTechnical information on classless routing�RFCs 1517, 1518, and 1519
�address assignment and aggregation strategy
�RFC1219�assignment of subnet numbers
�ftp://ftp.sesqui.net/pub/tools/tree.tar�program to help calculate address assignment
114
Paul Traina / INET '95Developing Countries Workshop
More informationTechnical information on address allocation
�RIPE NCC address allocation guidelines
115